This cross encoder performs sequence classification for contradiction/neutral/entailment labels. This has drop-in compatibility with comparable sentence transformers cross encoders.

I trained this model by initializaing the ModernBERT-large weights from the brilliant tasksource/ModernBERT-large-nli zero-shot classification model. Then I trained it with a batch size of 64 using the sentence-transformers AllNLI dataset.

For the large version, I froze all layers initialized from the tasksource model up to 19, and fine tuned only the remaining layers with a new classification head.

Features

High performing: Achieves 92.02% and 91.10% on MNLI mismatched and SNLI test.
Efficient architecture: Based on the ModernBERT-large design (395M parameters), offering faster inference speeds.
Extended context length: Processes sequences up to 8192 tokens, great for LLM output evals.

NLI Evaluation Results

F1-Micro scores (equivalent to accuracy) for each dataset. Performance was measured at bs=32 using a Nvidia Blackwell PRO 6000 Max-Q.

Model	finecat	mnli	mnli_mismatched	snli	anli_r1	anli_r2	anli_r3	wanli	lingnli	Throughput (samples/s)	Peak GPU Mem (MB)
`MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli`	0.8233	0.9121	0.9079	0.8898	0.7960	0.6830	0.6400	0.7700	0.8821	454.96	3250.44
`dleemiller/finecat-nli-l`	0.8227	0.9152	0.9265	0.9162	0.7480	0.5700	0.5433	0.7706	0.8742	539.04	1838.06
`tasksource/ModernBERT-large-nli`	0.7959	0.8983	0.9229	0.9188	0.7260	0.5110	0.4925	0.6978	0.8504	543.44	1838.06
`dleemiller/ModernCE-large-nli`	0.7811	0.9088	0.9205	0.9273	0.6630	0.4860	0.4408	0.6576	0.8566	540.74	1838.06
`cross-encoder/nli-deberta-v3-large`	0.7618	0.9019	0.9049	0.9220	0.5300	0.4170	0.3758	0.6548	0.8466	448.35	3250.44

Usage

To use ModernCE for NLI tasks, you can load the model with the Hugging Face sentence-transformers library:

python

from sentence_transformers import CrossEncoder

# Load ModernCE model
model = CrossEncoder("dleemiller/ModernCE-large-nli")

scores = model.predict([
    ('A man is eating pizza', 'A man eats something'),
    ('A black race car starts up in front of a crowd of people.', 'A man is driving down a lonely road.')
])

# Convert scores to labels
label_mapping = ['contradiction', 'entailment', 'neutral']
labels = [label_mapping[score_max] for score_max in scores.argmax(axis=1)]
# ['entailment', 'contradiction']

Training Details

Pretraining

We initialize the tasksource/ModernBERT-large weights.

Details:

Batch size: 64
Learning rate: 3e-4
Attention Dropout: attention dropout 0.1

Fine-Tuning

Fine-tuning was performed on the SBERT AllNLI.tsv.gz dataset.

Validation Results

The model achieved the following test set performance after fine-tuning:

MNLI Unmatched: 0.9202
SNLI: 0.9110

Model Card

Architecture: ModernBERT-large
Fine-Tuning Data: sentence-transformers - AllNLI.tsv.gz

Thank You

Thanks to the AnswerAI team for providing the ModernBERT models, and the Sentence Transformers team for their leadership in transformer encoder models. We also thank the tasksource team for their work on zeroshot encoder models.

Citation

If you use this model in your research, please cite:

bibtex

@misc{moderncenli2025,
  author = {Miller, D. Lee},
  title = {ModernCE NLI: An NLI cross encoder model},
  year = {2025},
  publisher = {Hugging Face Hub},
  url = {https://huggingface.co/dleemiller/ModernCE-large-nli},
}

License

This model is licensed under the MIT License.